Testposition 60 - Bishop vs Knight

[skoony]

samuel reshevsky?.(forgive my spelling)
mike

[Kyodai]

Well, actually I would go a couple of steps further with this position - i.e.
I would rather call it "a distinct or clear advantage" for black. The main
issue for the bishop's supremacy is of course pawns on both flanks.

I would be shocked if this game were forcibly won for Black, or even if were considered as a clear advantage for Black.
It would strongly violate the near-equality of Knight and Bishop in chess.

Robert

Yes Robert - we all have our frames of mind - but life is full of shocks...

Actually I found out that endgame expert GM Karsten Muller already has
investigated this famous ending Stoltz-Kashdan, with the discussion Bishop vs Knight - winning or not?

Karsten was convinced that white should be able to hold the position - with best defence.
But Burkhard Treiber from Germany finally managed to show how black wins this ending!
At least versus the variations given in Fundamental Chess Endings!

So the statement "clear advantage for black" is not too much here...

Again - The main issue for the bishop's supremacy, in this ending, is of course pawns on both flanks.

Link to the article "A perplexing riddle" in Chess Café

http://www.chesscafe.com/text/mueller109.pdf

[Houdini]

It's not about a personal "frame of mind" - I was thinking about the impact of this position on the end game play of chess engines (and Houdini in particular).
Look at the Houdini 3 analysis, at search depth 30 the position is still considered nearly equal, eval -0.18:

Code: Select all

30/55  18:52  -0.18    1...Kf8 2.Kf1 Ke7 3.Ke2 Kd6 4.Kd3 Kd5 
                       5.Kc3 h5 6.Nc2 g5 7.Nb4+ Ke4 8.Kc4 Be6+ 
                       9.Kc3 h4 10.Nd3 Kd5 11.Nb4+ Kd6 
                       12.Nc2 Ke5 13.Nb4 Bd7 14.Nd3+ Kd5 (6.593.200.089) 5819 

The search results shows that within the limits of the depth 30 search the "Knight equals Bishop" paradigm stands.
Will require some further study...

Robert

[Houdini]

OK, from what I can see after some analysis with Houdini 3.

After 12.Ng4? the game is definitely lost.

The first White error appears to be 9.Nc2?.
Houdini 3 would play the simple 9.b4! and continues to produce a 0.16 evaluation at search depth 30:

Code: Select all

30/52  12:35  -0.16    9.b4 Bb7 10.f3 f6 11.Nb5 a6 12.Nd4 h5 
                       13.Ne2 Ke5 14.Kd3 g5 15.Ke3 gxh4 
                       16.Nf4 Ba8 17.Nxh5 Bd5 18.Nf4 Bf7 
                       19.Ne2 Kd5 20.Kd3 Bg6+ 21.Kc3 Bh7 
                       22.Nd4 Ke5 (4.604.081.375) 6090

Even as late as move 12 the game could still have been saved by: 12.f3+ Kd4 13.Nc2 Kc5 14.Ne1! Kb4 15.Nc2+ Kxb3 16.Nd4+ Kb2 17.Nc6 Bf1 18.Nxa7 Bxg2 19.Ke2 securing the draw by allowing a trade of the b3 and g2 pawns for the a7 pawn. Black has won a pawn but not the game, White constructs a fortress.

Neither the simple 9.b4! nor the unexpected 14.Ne1! in the 12.f3 line were discussed in the ChessCafe article.

Robert

[Kyodai]

This is all very interesting! I will have a look at the variants when provided with a little more time

But I assure you - if we ask 20 GM:s "would you rather play this ending with the Knight or the Bishop - everybody would answer:
"With the bishop of course"
Aha and why is that?
"Bishop is superior to Knight in this type of ending with pawns on both flanks" that's basic knowledge.
Aha - so if we just keep the pawns on the Kingside?
"Then it doesn't matter - I can play with Knight or Bishop - it's anyway dead draw"

So clear advantage for black in this position - means only black has real winning chances.

From research (if I understand correctly) the ball is on white's side i.e. the
latest variations (according to GM Muller) show black is winning.

So the whole discussion is of course very interesting.

Don't know how to specific bring this knowledge into chess engines - but
the criterias are

1) Clean cut ending with one bishop vs one knight

2) No double pawns

3) Pawns on both queen-and kingside

4) "Normal" king positions i.e. no side has a superior king

With these criterias engines should give a significant plus score to the side
with the bishop - the master can tell just by a swift look at the position.

[zullil]

With these criterias engines should give a significant plus score to the side
with the bishop - the master can tell just by a swift look at the position.

The true score of the position is -1, 0 or 1 (ie, loss, draw, win). If the position is a draw with best play, then assigning a significant plus score to either side is simply wrong. Based on the analysis posted already by Robert H., I'm inclined to think this position is a draw. But I'm a weak player.

By the way, here's Critter's analysis at this point:

Code: Select all

31/56  37:42   23770M  10508K  +0.15 Kf8 Kf1 Ke7 Ke2 Kd6 Kd2 f6 Kc3 Kd5 b4 a6 h4 h6 Nf3 g6 Ne1 Bb5 Nc2 h5 Nd4 Ke5 Nf3+ Kf4 Nd4 Bf1 g3+ Ke5 Nb3 Bb5 Nd2 Kd5 Nb1 Bd7 Nd2 Bc6 Nb3 g5 Nd4 Bb5 Nf3 Ke4 Nd2+ Kd5 Nf3

[peter]

Hi Robert!

Even as late as move 12 the game could still have been saved by: 12.f3+ Kd4 13.Nc2 Kc5 14.Ne1! Kb4 15.Nc2+ Kxb3 16.Nd4+ Kb2 17.Nc6 Bf1 18.Nxa7 Bxg2 19.Ke2 securing the draw by allowing a trade of the b3 and g2 pawns for the a7 pawn. Black has won a pawn but not the game, White constructs a fortress.

I don't think the game can be saved by 12.f3+, because 13...Kc5?! in your variant seems not as strong as 13...Ke5! to me:

1... Kf8 2. Kf1 Ke7 3. Ke2 Kd6 4. Kd3 Kd5 5. h4 Bc8 6. Nf3 Ba6+ 7. Kc3 h6 8. Nd4 g6 9. Nc2 Ke4 10. Ne3 f5 11. Kd2 f4 12. f3+ Kd4 13. Nc2+ Ke5 14. Nb4 Bb5 15. Nd3+ Bxd3 16. Kxd3 Kd5 17. b4 Kc6 18. Kc4 Kb6 19. Kd4 Kb5 20. Kc3 Ka4 21. Kc4 Ka3 22. Kc3 Ka2 23. b5 Ka3 24. Kc4 Kb2
0-1

[Houdini]

I don't think the game can be saved by 12.f3+, because 13...Kc5?! in your variant seems not as strong as 13...Ke5! to me:

1... Kf8 2. Kf1 Ke7 3. Ke2 Kd6 4. Kd3 Kd5 5. h4 Bc8 6. Nf3 Ba6+ 7. Kc3 h6 8. Nd4 g6 9. Nc2 Ke4 10. Ne3 f5 11. Kd2 f4 12. f3+ Kd4 13. Nc2+ Ke5 14. Nb4 Bb5 15. Nd3+ Bxd3 16. Kxd3 Kd5 17. b4 Kc6 18. Kc4 Kb6 19. Kd4 Kb5 20. Kc3 Ka4 21. Kc4 Ka3 22. Kc3 Ka2 23. b5 Ka3 24. Kc4 Kb2
0-1

I used 13...Kc5 from the ChessCafe analysis, but indeed Ke5 could be stronger.
The pawn ending after 15.Nd3+ Bxd3 is lost indeed, maybe 15.Kc3 saves the day?
Interesting stuff .

Whatever the conclusion of this line, White's suffering is self-inflicted after 9.Nc2?. There's no reason to give the black King a free passage to e4.

[peter]

I used 13...Kc5 from the ChessCafe analysis, but indeed Ke5 could be stronger.
The pawn ending after 15.Nd3+ Bxd3 is lost indeed, maybe 15.Kc3 saves the day?
Interesting stuff .

Whatever the conclusion of this line, White's suffering is self-inflicted after 9.Nc2?. There's no reason to give the black King a free passage to e4.

Of course the single variant above is not the only one to be considered.
I don't dare finally judging the starting position till now, it's really time consuming to keep all relevant points in hash at backward solving.
But as for the first 11 moves I'd say White is lost afterwards.
E.g. at your 9.b4 I'll have to look closer.
Here's just a little .pgn from what I have till now:

[Event "Bishop Superiority"]
[Site "The Hague"]
[Date "1928.??.??"]
[Round "?"]
[White "Stoltz"]
[Black "Kashdan"]
[Result "0-1"]
[Annotator "Larsson"]
[SetUp "1"]
[FEN "6k1/p2b1ppp/8/8/3N4/1P5P/5PP1/6K1 b - - 0 1"]
[PlyCount "39"]
[EventDate "1928.??.??"]

{This position is from Stoltz-Kashdan 1928. A celebrated ending where the
existence of pawns on both wings on an open board emphasizes the bishop's
superiority over the knight.} 1... Kf8 2. Kf1 Ke7 3. Ke2 Kd6 4. Kd3 Kd5 5. h4
Bc8 {Black's bishop helps push the white king backwards, enhancing the scope
of his own king.} 6. Nf3 Ba6+ 7. Kc3 h6 8. Nd4 g6 9. Nc2 Ke4 10. Ne3 f5 11. Kd2
f4 12. Ng4 (12. f3+ Kd4 13. Nc2+ Ke5 14. Nb4 Bb5 15. Nd3+ (15. Kc3 a5 16. Nc2 (
16. Nd3+ Bxd3 17. Kxd3 Kd5 18. Kc3 Kc5 19. Kd3 Kb4 20. Kc2 Ka3 21. Kc3 h5 22.
Kc2 Ka2 23. Kc3 Kb1 24. Kc4 (24. Kd4 Kb2) 24... Kc2) 16... Bf1) 15... Bxd3 16.
Kxd3 Kd5 17. b4 Kc6 18. Kc4 Kb6 19. Kd4 Kb5 20. Kc3 Ka4 21. Kc4 Ka3 22. Kc3 Ka2
23. b5 Ka3 24. Kc4 Kb2) (12. Nc2 Bf1 13. f3+ Ke5 14. Ne1 h5 15. Nd3+ Kf5 16. b4
g5 17. hxg5 Kxg5 18. Nc5 Kf5 19. Ke1 Bxg2 20. Kf2 Bh3 21. Nd3 Kg5 22. Ne5 a6
23. Nd3 Be6 24. Kg2 Bc4 25. Nc5) 12... h5 13. Nf6+ (13. Nh6 Bf1 14. f3+ Kd4 15.
Nf7 Bxg2 16. Ng5 Bf1 17. Ne6+ Ke5) (13. f3+ Kf5 14. Nf2 g5 15. hxg5 Bf1) 13...
Kf5 14. Nd7 (14. Nh7 Kg4 15. Nf8 Kxh4 16. Nxg6+ Kg5) 14... Bc8 (14... Bf1) 15.
Nf8 (15. Nc5 Kg4 16. Ke1 Kxh4 17. b4 Kg4) 15... g5 16. g3 gxh4 (16... Kg4) 17.
gxh4 Kg4 18. Ng6 Bf5 19. Ne7 Be6 20. b4 Kxh4 {and the passed h-pawn carried
the day.} 0-1

And yes, this endgame is hot

[Karlo Bala]

This is merely a discussion (and a celebrated game!) of B vs Kn in an ending with pawns on both flanks.
Such an ending is very difficult to hold for white

Without looking at the game or at any engine analysis, I find it hard to believe that this position would be lost for White.
Maybe Black has a very slight advantage with the bishop against knight, but I cannot believe it's enough for the win.

Well, actually I would go a couple of steps further with this position - i.e.
I would rather call it "a distinct or clear advantage" for black. The main
issue for the bishop's supremacy is of course pawns on both flanks.

Yesterday I looked around a bit to see if this game was carefully commented somewhere. But couldn't find anything that could cast further
light over this issue (won for black or not?)

Merely a feeling - but I don't expect any GM to really state that this is
a winning position for black - more like it practically is very very difficult to defend for white. But I don't know - will check up with a GM of elo 2580,
maybe also with endgame specialist Karsten Muller.

IMO, the fact that there are pawns on both sides is not enough for black to win. Black has at least the two more advantages. Black has a tempo, and white knight is bad placed (ideally, on d4 should be a white king). Move the white knight anywhere else and give white a tempo, and position is a dead draw.